BackgroundFew longitudinal studies exist to guide clinicians or administrators on what functional outcomes can be expected among nursing home residents with different levels of cognitive impairment. Extrapolating from cross-sectional studies or from longitudinal studies of community-dwelling residents may provide misleading estimates of prognosis, hindering efforts to target preventive care.ObjectiveTo describe patterns of change in physical function on a quarterly basis over 1 year among long-stay nursing home residents grouped according to their level of cognitive impairment on admission.MethodRetrospective analysis of activities of daily living dependence ratings were based on quarterly MDS+ assessments from 76,016 long-stay residents admitted to nursing homes during calendar years 1993 through 1996 in five states participating in the National Case Mix and Quality and Demonstration Project. Residents were stratified by level of cognitive impairment on admission using a 7-level Cognitive Performance Scale. The activities of daily living dependence was measured by a 20 point scale. Mean activities of daily living scores on admission to the hospital and at four quarterly intervals following admission were compared across cognitive impairment levels and by state of residence.ResultsA change in activities of daily living dependence over 1 year in most groups averaged 1 point or less. Three patterns of activities of daily living dependence were identified consistently across five states. Those with mild cognitive impairment on admission showed an initial reduction in dependence followed by slow increase; those with moderately severe impairment showed slow linear increased dependence; and those with severe cognitive impairment showed an initial improvement in dependence, followed by stability.ConclusionMore complex statistical models that take into account comorbid conditions at baseline, in addition to cognitive performance, might identify subgroups of nursing home residents who are at risk for rapid decline. Ways to better characterize declines in function are needed, otherwise relatively large samples will be required for intervention trials.